One question, carried all the way · step 7 of 8
Planning the follow-up
Should we repeat this, when does it go stale, and what next?
A decision made once is a decision that decays. The same design machinery that sized the first trial says when this answer stops being usable and what the next dose worth testing is.
The answer is not 'run it again annually because that is the budget cycle'. It is a number: how long before the value of a fresh answer exceeds what a fresh answer costs.
step 7
from axiom.design import time_to_re_experiment
timing = time_to_re_experiment(
posterior_sd=float(after.summary.sd) / REGION_DEPOTS,
half_life_periods=26.0,
experiment_se=se_experiment,
min_eig=0.5,
)
print(f"a repeat is worth running again after {timing.periods:.0f} weeks")
print(f" today it would gain {timing.eig_now:.3f} nats, "
f"against a bar of 0.5")
what it printed
a repeat is worth running again after 39 weeks today it would gain 0.239 nats, against a bar of 0.5
Every line above was captured by running this code at build time. The steps share one namespace, so the code on this page is the code you would type after the steps before it.